CPU microcode is like a mini BIOS in the CPU. Usually processor manufacturers will send out microcode updates to motherboard manufacturers who will include it in the mobo BIOS updates. When the system is booted after the mobo BIOS update, it will update the CPU microcode. Quite a lot of CPU bugs we're fixed this way, mostly those dealing with calculation accuracy.

USM student (& also screw up PCs during my free time ) But I dunno how to program bios I read about that in an old book on PCs. Published back in the day when the first K8 processors came out (socket 940, not 939).

My AM2 OC result is long overdue..been too busy tweaking my legacy setup (built for win 98 applications). Managed 2.7GHz on my Windsor F2, but not stable at all (not enough cooling to pump enough voltage). Gonna work on it after exams, anyway my AM2 is always at home while I'm always at campus except during holidays, its destined to be my HTPC/home server eventually.

I'm an applied physics student, staying at the new hostel at the back of USM (across yeah chor ee road). Give me a call (016-4936208) if you wanna meet me, bt my am2 set is at home, not in my hostel.

Damn, I 1 week not online tambah 7-9 pages..read until pening

No prob..I memang got no harapan to have the best PC..what I want is to be very inovative in making the most out of what I can get..

I've long since lost my trust in any board-based voltage monitoring, whether its bios, everest, speedfan or any other program..whenever I take out my digital multimeter and test the board directly I get very different results from what I see in software. Either the value is outright wrong, it fluctuates when its supposed to be fairly constant. Even if most of the values are vaguely correct, there is always at least one which is completely screwed up. Can't find any reason except that the monitoring chip on the motherboard has f***ed accuracy.

Not really, depends on the PSU design. Typically most (80-90%) of reliable PSUs have good efficiency (75-80%) from 20% load to 80% load. Below 20% and above 80%, efficiency is typically worse, around 65-75% Which means,

450W PSU
Efficient range 90-360W

600W PSU
Efficient range 120-480W

1000W
Efficient range 200W-800W

Of course this may vary, highly depends on the design of the PSU. Some rare ones have good efficienty (70-85%) at even very low (=<10%) and very high load (90-120% -120 is overloaded) while some have abyssmally bad (<60%, I've even seen one as bad as 40 something %) efficiency at those high and especially low loads. To be sure, check out reviews by jonnyguru and silentpcreview, they record the efficienty of PSUs thoughout the operating range.

I have a friend who has a Kill-a-watt meter, and another meter which is slighly better. I haven't tried either yet, currently I'm just using my digital multimeter's AC current measurement function to measure my PC'sower consumption. One day I'll borrow his meters and make a comparision, if possible with an osciloscope as well (have to drag PC to physics lab; heavy le..).

Thats really nice, but damn its out of my reach (financially)

Those compounds will help, but only a few degrees at most. From best to worst, Artic silver ceramique, Artic silver 2, thermalright supplied compound.

---

Added on October 27, 2007, 6:39 pmDamn, just read through the Black edition review..I want a brisbane G2 for my HTPC (must be cheap la)..no need fast to hold any performance records..even at 3.1GHz the power consumption is damn low..can passive cool if necessary..

---

Added on October 27, 2007, 7:00 pmFound out how to idenify the new Brisbane G2 processors, OPN ends with DO.

Someone faster find the price..

This post has been edited by lohwenli: Oct 27 2007, 07:00 PM

That tomshardware review of the 5000 Black Edition counts as a Brisbane G2 review. Read it in full detail. G2 has improved overclockability and greatly lowered power consumtion even at high frequencies. 3.0GHz on stock voltage should be possible for most G2, and if the voltage is kept that low it will only take 7* watts at most, which allows it to compete with C2D even in full load power consumption and performance (it pulls up even with the E6550 when OC to 3.1GHz). Things are looking good for AMD's 65nm processors, even Barcelona has a new stepping, B2 (original release is A0), hopefully it might resolve the clockspeed issues that plagued the original Barcelona release.

Btw, before you fire me about the known fact that Brisbane is slower than Windsor, there hasn't been any comparision between the G1 and G2 stepping in terms of clock for clock performance. In the tomshardware review though, at 3.1GHz it beats the 3.0Ghz Windsor F3 by roughly 1.5% in non-cache size sensitive applications with a 3.33% difference in clock speed(in cache sensitive applications it gets pawned by the 6000's and the E6550 large caches;1MBx2 and 4MB respectively). So yes, its most likely it will never beat the Windsor F3s in overclocked performance, but it now can be clocked up to high enough speeds while remaining really cool.

Yes, I doubt G2 can possibly win F3 for now. But There's one thing, at 3.0GHz a F3 is putting out 100W+ while a G2 is only 7*W. Technically its possible to reach 3+GHz on stock heatsink, which makes it an absurdly easy chip to overclock. No need fancy cooling or even heavy PSU.

Congrats bro..
If only K8 can match C2D on clock for clock performance, you'll be a serious contender with C2D overclockers.. Hopefully K10 will also be able to clock up high.

The thing is, though it maxes out similar to brisbane, reaching 3.0Ghz is now dirt easy even with the stock cooler..
In fact you could even slow down the fan and it'll still be cool enough to be stable..

Sadly, I think your statement is likely to turn out correct. Just saw tomshardware's review of their Penryn engineering samples, Phenom will have to pull out a miracle (like clock 4+Ghz)to beat it. There might be some hope with Deneb (45nm version of Agena), which will be out in the middle to end of next year. Shanghai (45nm of Barcelona) will be out within the first half of next year.

Yup, one of the cores is disabled. Official statement by AMD, though I can't find the page.


Wait for the next architecture - Bulldozer.

Forgot about that..yup, its one of the really cool features of Phenom. And unfortunately it will be sorely needed to compete with Penryn where power consumption is concerned (the quad core Penryn uses only 7x watts-measured by tomshardware).

This post has been edited by lohwenli: Nov 5 2007, 12:09 AM

Its not the black edition that I'm looking out for, but for cheap brisbane G2. The black edition is really out of my budget (economically, and common sense wise). And unfortunately I also missed out on your lovely F3s. But then again, a G2 for me is more suitable for where its going to end up-inside my HTPC (an overclocked and underclocked one at that).

Thats if I can find any at a price I can afford (around RM250 or less). I doubt i can beat F3 in performance even if its a miraculous overclocker, it'll need to clock 4.2GHz to have any chance of beating F3. But it might still overclock better than 3.2GHz, tomshardware isn't well known for pushing their procs to the max.

But that worries me..they don't push to the max and yet penryn can clock 4ghz on aircooling.. AMD, hurry up with Deneb already..

It should be. I seriously doubt it will outperform a overclocked C2Q even when its overclocked. I suspect it will only clock up to 3.x at most, probably not even as good as the F3 Windsors. I hope they will improve from the Barcelona A0 (or was it A1?), the B* stepping seems to be better but found no overclocking results yet. Hope Phenom will launch on a C* stepping or better.

Oh yeah, those of you who love memory benchmarks and superpi records will love phenom..can easily pwn C2Q like kacang putih only..this is from Barcelona results, Phenom should be very similar.

It was true for the first stepping, A0, of barcelona, 2.4GHz also damn hard to achieve. But according to AMD its also because the dual processor setup (8 core total) affects the max stable clock speed. The current stepping is B0 (not sure about its overclocking results), BA and B2 revisions wll be available next month, which are anticipated to allow barcelona to reach 3.0GHz.

Just read the news reports, seems Phenom is also clockspeed limited (up to 2.4-2.5GHz)for the initial stepping, which is B0. There was supposed to be a 2.6GHz model, but seems they had the same problems as barcelona. Hopefully Phenom will update to a better stepping quickly. Price looks interesting though, the X4 is roughly the same price as current mainstream (non-high end) 65nm C2Q. But means it will still be out of my reach for prehaps another half year. Probably I'll skip straight to Deneb.

Adding salt will allow the ice to melt a few degrees below zero, which will allow for cooler melt-off to reach your processor (below 0 water is ice, which won't reach your processor). Adding other substances won't help much, you can only drop the melting point of ice be a few degrees at most by adding stuff to it.

Here's one way you can go lower on your current WC set with minimal effort (though like ice its only for a short time). Run alcohol (methylated spirit is an easy source) instead of water in your WC set. Then you can use dry ice to chill it (below zero if necessary, that's why you can't use water, it'll freeze). Control the temp by the amount of dry ice, or you may hve condensation problems. Also, if you're tempted to go overboard and run an extreme subzero run, be careful that sudden temp changes can cause the plastic parts of waterblocks to crack, especially when its already at very cold temps (plastic becomes more brittle the colder it gets).

I'm not a chemistry guy, I'm a physics guy, but I hope what I said above helps.


Only peltiers, phase change, dry ice and LN2 can get you below zero C. I have a possible subzero experiment planned for during my year end hols, I'll update when I get around to doing it.

You don't have to use so much dry ice until the temp drops so much. Just enough till you achieve the desired temp. Up til around -10 to -20 should not be a problem. Its -50 and below that would be worrying, but by then condensation would have hit you first.

Forgot that you're in Penang..my exams end in a weeks time, then perhaps I can meet you and we can have some OC fun..perhaps even do a little test run of my planned experiment..

If the temp is low enough, condensation will happen no matter what unless you're running in a moisture free environment. Doing what you said will help, but only up till a certain temp.

Putting dessicants will help, but as long as there is even 1% moisture there will be condensation at some point. However the temp range where water remains in liquid form will become very small, as the dew point will be close to the freezing point; few degrees more and become ice already (which is non conductive).

Its normal for voltage to drop during load, most motherboard power regulators will suffer from some vdroop, forcing OCers to need to set higher voltages than usual. Btw, if your processor voltage is fluctuating, don't blame the PSU first, the power regulator for the processor is on the motherboard, not the PSU.


If using duck brand PSU, more vcore will cause the duck to get roasted..
Cos processor now sucking more power from 12v rail..

It's weird no, often the high power consuming chips overclock well, like the Intel Netburst based processors and the HD2900.

Hard to say, lets do the math

Estimate your PC uses 150W(actually less, I run almost the same system as yours, and I run just over 100W on full load OC to 2.3GHz), and another 70W for a CRT monitor.

220W x 24hours x 30days = 158.4 KWH
1KWH is more or less 0.25 sen, so total is RM39.60 to run the PC at full load with monitor on 24/7.

Running on onboard graphics, its unlikely your PC would cost more than RM50 to run 24/7. With single GFX, then the cost goes up another 25-100%, depending on what card (also assuming the card is under load for at least 50% of the time). SLi/crossfire is what would really break the bank, I'm not sure even how to estimate for that.

1.5v Vcore should not be a problem for Windsor F2 if cooling is good enough. The Athlon FX that uses Windsor F2 runs on a stock vcore of 1.4-1.45v, it does come with a heatpiped cooler though.

This post has been edited by lohwenli: Nov 17 2007, 12:33 AM

Depends on processor, if there is no divider that will give exactly 800 while on stock CPU speed then it will run slightly below 800 unless you overclock the processor.

F3 can go waaaay higher..even better than Brisbane G1..
Ram go get anything using D9GMH, D9GKX chips. D9GCT also ok if you're on a budget.

Phenom loses in gaming and some multimedia apps due to the smaller cache ( L1, L2 and L3 combined). Lucky the K10 memory controller is a real killer, thats why its hitting some murderous superpi records (when running a dual/octa barcelona, each processor counts as an additional dual channel DDR2, so you can get 4-16 channel memory access-my god the memory benchmark looks more like a cache benchmark ). Unfortunately for a single processor this advantage is minimal, and not enough to offset the lack of cache (which is even worse when compared with Penryn).

Which brings me thinking. How badly wil a Q9xxx get pawned by a Phenom FX dual processor system? Since there will be more cache (though not really shared), and the memory bandwidth will increase drastically due to a 2nd dual channel controller, the weakness may be overcome (also assuming that the programs won't actually use all 8 cores, to be realistic).

Its true, reported on Tom's Hardware Daily, the Inquirer, Digitimes and a few others. At least one includes an official statement by AMD stating that there are availability issues. Hope its got nothing to do with low yield of high clockable chips, that would be very bad news for overclockers.

Impressive overclocking (3Ghz) on stock voltage (1.225v), not as good as Windsor F3, but defintely pwns Brisbane big time. Unfortuntately its still not clocking fast enough to pull up on par with C2Q, more so the lack of cache really hurts in gaming and multimedia. But the price is Some more can just drop into any AM2 board, its a damn simple upgrade compared with any Intel system (which will be obsolete in a year anyway), and the rest of the components are cheaper too Even Crossfire/SLI is easier, and there are fewer trade offs.

Fyi, 754/939 stock hsf also fits perfectly on AM2.

There's something missed out in this review, the ram is running at 5-5-5-15. Everyone knows that AM2 at that timings can barely outperform an equivalent 939 in any memory/cache related task. Memory latency is much much more important than bandwidth for K8/Stars processors because of the low latency advantage of the integrated memory controller, which is needed to offset the disadvantage of its pathetic cache size.

This post has been edited by lohwenli: Nov 22 2007, 12:19 AM